Are Debt Sustainability Indicators Based on Time-Series Data Useful for Predicting Crises?

Abstract

A large literature in empirical public finance applies time-series techniques to historical data and draws inference about public debt sustainability of individual countries. These methods include unit-root tests on primary deficits and cointegration between revenue and expenditure, as well as fiscal reaction functions. In this note, we take a systematic approach to evaluating the in- and out-of-sample performance of various methods in predicting sovereign debt crises. In a panel-logit regression analysis for 31 countries, we find that the benefits for forecasting are surprisingly small.